SEAVEAtk

A range of software toolkits are available that address various facets of UQ. The present team of investigators has, over the past three years, built on this work, and honed it to provide a powerful, interoperable, open source and open development environment into which existing and new UQ capabilities can be inserted. These tools are intended to be general-purpose, cross-cutting, and highly flexible, our priority demonstration applications cover fusion energy, weather, biomedicine, advanced materials, human migration, and geosciences. Ease of use is a central design feature, by which existing applications can be quickly instrumented with VVUQ software, and rapidly deployed on supercomputers. As a testament to these capabilities, several of our VVUQ software items have been downloaded thousands of times.

Will establish the SEAVEA toolkit which aims to ensure that we can orchestrate full VVUQ on the NEPTUNE code. Official releases will be in M12 and M24 and a final release in M30, with the remaining months used to promote and harden the full version.

The SEAVEA toolkit establishes a platform for verification, validation and uncertainty quantification (VVUQ), building on the work VECMAtk and UQ codes from the Alan Turing Institute (ATI). The goal is to provide tools that can be combined to capture complex scenarios, applied to applications in disparate domains, and used to run multiscale simulations on any desktop, cluster or supercomputing platform.

Month 12 Release of SEAVEAtk

Figure 1: Graphical overview of the components in the SEAVEA toolkit release and their integrations.

The M12 release is made up of the following tools:

• EasyVVUQ
• QCG PilotJob
• EasySurrogate
• FabSim3
• Muscle3
• MOGP Emulator

EasyVVUQ v1.1.2

• Documentation: https://easyvvuq.readthedocs.io/en/latest/
• Code: http://www.github.com/UCL-CCS/EasyVVUQ

Updates

Overhaul of SC sampler / analysis class:

• Made isotropic sparse-grid subroutines more scalable to higher input dimensions. Reused dimension-adaptive subroutines for this purpose, instead of having (slower) separate isotropic routines.
• Rewrote dimension-adaptive SC expansion as a standard PCE expansion with generalized PCE coefficients. See adaptive sparse-grid tutorial.

Documentation

• Extensive methodological sparse-grid tutorial: https://www.researchgate.net/publication/359296270_Adaptive_sparse-grid_tutorial
• New tutorial on using mathematical expressions involving parameters in template files using the Jinja encoder: https://github.com/UCL-CCS/EasyVVUQ/blob/dev/tutorials/jinja_tutorial.ipynb

QCG PilotJob v0.13.1

• Documentation: https://qcg-pilotjob.readthedocs.io/en/latest/
• Code: https://github.com/vecma-project/QCG-PilotJob

New features

• Additonal environment variables for launching srun jobs from inside a pilot job parallel task

Bugfixes

• Support for different argument names between Slurm versions (Lourense Veen)
• Race condition in wait4all method (LourenseVeen)
• Reporting version in debug logs

Other Improvements

• Enhanced tests (Lourense Veen)
  • support for tests with different Slurm versions
  • updated documentation for tests
  • bugfixes within the testing scripts

EasySurrogate v0.19

• Tutorial: https://github.com/wedeling/EasySurrogate/tree/master/tutorials
• Code: https://github.com/wedeling/EasySurrogate

Updates

Some minor changes have been introduced in the ANN / Deep active subspace methods:

• made it possible to select varying numbers of neurons, to create “constrained” hidden layers. This is used to make a deep-active subspace type of surrogate without orthonormality embedded in network architecture. Instead, orthonormality is computed in a post-processing procedure. Explained in updated tutorials.

Tutorials

• 2 epidemiological deep-active subspace Jupyter notebook tutorials (HIV and COVID 19): https://github.com/wedeling/deep_active_subspace_data

FabSim3 v3.2

• Documentation: https://fabsim3.readthedocs.io/en/latest/
• Code: http://www.github.com/djgroen/FabSim3

New Features

• Two new plugins: FabSCEMa and FabNEPTUNE
• Range of fixes and improvements to the FabMOGP plugin
• Inclusion of the FabSim3 Automated Validation tutorial.
• Added support for multiplexing in the SSH layer, and tutorial documenting it. This greatly reduces the number of times users have to type in passwords.
• (We also submitted a paper describing FabSim3 to a journal)

Updates

• A range of improvements to the documentation.

Tutorials

This version of FabSim3 has been used to enable the following application tutorials:

• • FabFlee (migration modelling) link to tutorial
  • FabMD (molecular dynamics) link to tutorial
  • FabCovid (Covid-19 simulation) link to tutorial
  • FabUQCampaign (basic CFD models) link to tutorial
  • fabmogp (Multi-Output Gaussian Process Emulator) link to tutorial

Muscle3 v0.5.0 (made possible through the Dutch e-Science Center)

• Documentation: http://muscle3.readthedocs.io
• Code: https://github.com/multiscale/muscle3/releases/tag/0.5.0

New Features

• MUSCLE3 now starts submodels and other components (using QCG-PilotJob)
• Automatic resource management for components on HPC

Incompatible changes

• compute_elements are not called components in .ymmsl files

Improved

• Integration with QCG-PilotJob
• Extension of yMMSL language
• Simplification of the installation process by removal of gRPC dependency and automatic detection of MPI availability
• Use latest OpenSSL library when installing it automatically
• Cleaner logging output
• Improved communication performance and scalability.

MOGP Emulator v0.7.0

• Documentation: https://github.com/alan-turing-institute/mogp-emulator/tree/main/mogp_emulator/demos
• Code: https://github.com/alan-turing-institute/mogp-emulator/tree/main/mogp_emulator

New Features

New release incorporating recent work. New features in this release:

• Maxi-min LHC Experimental Designs
• Mulit-output history matching
• Emulator validation
• GPU implementation now supports Priors
• Additional fixes and improvements to documentation and examples

Previous Releases

SEAVEAtk V0.1

EasyVVUQ v1.1.1

• Documentation: https://easyvvuq.readthedocs.io/en/latest/
• Code: http://www.github.com/UCL-CCS/EasyVVUQ

QCG PilotJob v0.13.0

• Documentation: https://qcg-pilotjob.readthedocs.io/en/latest/
• Code: https://github.com/vecma-project/QCG-PilotJob

EasySurrogate v0.18

• Tutorial: https://github.com/wedeling/EasySurrogate/tree/master/tutorials
• Code: https://github.com/wedeling/EasySurrogate

FabSim3 v3.6

• Documentation: https://fabsim3.readthedocs.io/en/latest/
• Code: http://www.github.com/djgroen/FabSim3

EQI (formerly EasyVVUQ-QCGPilotJob) v0.5.2

• Documentation: https://easyvvuq-qcgpj.readthedocs.io/en/latest/
• Code: https://github.com/vecma-project/easyvvuq-qcgpj

MUSCLE 3 v0.4.0

• Documentation: http://muscle3.readthedocs.io
• Code: https://github.com/multiscale/muscle3/releases/tag/0.4.0

MOGP Emulator v0.6.1

• Documentation: https://github.com/alan-turing-institute/mogp-emulator/tree/main/mogp_emulator/demos
• Code: https://github.com/alan-turing-institute/mogp-emulator/tree/main/mogp_emulator